The quote on this subject in the blog post is downright offensive...

Make the list bigger!

Indeed! AI slop isn't inevitable, it's just so _easy_.

But here's my high-effort poem, written just now by GPT 5 Thinking, with 20 turns of conversation for me to get it started and mostly to refine the result. I think it's pretty good. (Text in alt.)

Life cannot be contained

Here's a paper about what happens to the evaporated water: essd.copernicus.org/articles/12/...

And a map from that showing what portion of evaporation is recycled into rain over land.

In general 70% of evaporated water returns as rain over land, though maybe only 5-15% in the same region.

Results so far for bounding box and segmentation masks on Gemini are quite disappointing. Gemini Pro can sometimes produce a competent bounding box, but Flash cannot. Neither can produce valid image masks.

(First example is Pro, then Flash)

The many hues present on Claude's update status page is... indicative of the actual experience, unfortunately.

I was talking about AI with my daughter, and mentioned it lying about doing tasks, and she was insistent that AI couldn't lie. Which of course it can...

I tried doing a test in chatgpt but was admonished that it "can’t provide intentional misinformation". But firing up the playground...

I made ChatGPT uncomfortable in the process... whether that makes it better or worse I'm not sure

I don't know exactly what you're looking for, as a professional... but it's better at horny squid verse than I am, as an amateur...

Thinking a little bit about the graph view from @wattenberger.com's Fish Eyes post: wattenberger.com/thoughts/fis... ... and specifically this concept graph. The concept graph is kind of like wandering Wikipedia, but happening in-place. It got me thinking of a more linear approach 🧵

Was literally just making this...